Theme: Nonlinear Optimisation
Course Title: Optimisation Techniques For Data Analysis

Lecturer: Professor Stephen Wright

Course Content
Optimisation techniques have become a mainstay of data analysis and machine learning. Many problems in these domains can be formulated and solved naturally as optimisation problems. The explosion of interest in data applications has led to renewed focus on optimisation techniques that are relevant to this area, and thus to re-examination and enhancement of many techniques that were in some cases previously thought to be of limited interest. Many important discoveries have been made over the past 5-8 years about the properties of such fundamental approaches as first-order methods, accelerated gradient, stochastic gradient, coordinate descent, and augmented Lagrangian techniques. Advances have been made too in the sophisticated implementation of these techniques to key problems in data analysis, and in their parallel implementation. This course will review basic optimisation techniques, their application data analysis problems, and their fundamental theoretical properties.

Background Reading

  • J. Nocedal and S. J. Wright, “Numerical Optimization” (2nd Edition), Springer, 2006.